人工神经网络在预测胰腺炎病程变化中的应用

目的]通过胰腺炎发生过程中酶学的时限变化建立人工神经网络(Artif-icial Neural Networks,ANN)模型来预测胰腺炎患者病程变化。方法]分析131例胰腺炎患者的临床资料及其血尿淀粉酶、血脂肪酶的测定时间和结果,把测定时间(h)、血尿淀粉酶、血脂肪酶测定值归为4个集合,应用MATLAB6.5软件描出拟合曲线,并建立胰腺炎发病全程酶学监测的ANN模型,同时随机抽取14例临床确诊患者(具有确切病程)的数据作为输入值进行模型验证,以检测网络的稳定性。结果]在发病的早期胰腺炎患者血脂肪酶和血淀粉酶基本上呈正相关(Y=24.5174 47.7886X,r=0.5282,P<0.01),而整个病程中血脂肪酶能更有效地反映患者病情变化,血尿淀粉酶和血脂肪酶拟合的ANN模型对胰腺炎病程预测的准确度为56.6%。结论]利用胰腺炎酶学ANN进行胰腺炎病程预测能起到良好效果。

The Application of artificial neural network in pancreatitis complication prediction

Objective To identify pancreatitis complication tendency development through artificial neural network science calculation.MethodsThe clinical data and laboratory data including urinary amylase, blood amylase,lipidase and testing time of one hundred and thirty-one pancreatitis cases was analysed. The data was divided into four groups according to the testing time. MATLAB6.5 software was used to depicture the fitting curve and the ANN model to monitor the enzymology changes of the whole range of pancreatitis was established. And fourteen cases of clinical definite pancreatitis were collected to test artificial neural network at the same time.Results In the early stage of pancreatitis, the blood lipidase and blood amylase had positive correlation(Y=24.517 4 47.788 6X,r=0.5282,P<0.01). During the whole course of pancreatitis, the blood lipidase could reflect the changes of pathogenetic condition. The accuracy rating of ANN model to predict the course of pancreatitis was 56.6%.Conclusion The ANN model to monitor the enzymology changes of the whole range of pancreatitis is a good method to predict the changes of pathogenetic condition.